First Steps Towards Federated Learning Network Traffic Detection

Federated Learning (FL) has emerged as a promising machine learning approach which enables cooperative training while preserving data privacy. The FL process involves clients and a parameter server to exchange model parameters over the network, generating unique traffic patterns that can be detected. The identification of FL traffic may be used in diverse applications, ranging from network traffic management to security and privacy enhancement in FL environments. For this purpose, this paper investigates the problem of FL traffic identification. We created a dataset of FL network traces, encompassing different FL training setups. We then developed two classifiers based on Random Forest and Neural Network algorithms and compared their performance, obtaining promising results.